Abstract: The high-order differential feedback controller is a control scheme inddepent of the precise model of the system. It indirectly compensates for the system unknown model function through a control filter. However, the high-order differentiator has not been used to estimate the system unknown model function, which is indirectly compensated. In this paper, an improved high-order differentiator is proposed by introducing control input information, which can estimate the unknown functions in nonlinear models and the differential information of system output and reference input in real-time. The convergence of the unknown model functions estimated by the improved high-order differentiator and the extended state observer is compared and analyzed. It is proved that the former is one type higher than the latter. A novel high-order differential feedback control algorithm is designed using the estimated differential and model functions, suppressing unknown disturbance and successfully controlling the Quadrotor Unmanned Aerial Vehicle(QUAV) attitude system. Lyapunov-based stability analysis is conducted to prove the stability of the closed-loop system. In the control test platform experiment based on the Pixhawk, schemes of the proposed improved high-order differential feedback control, PID control, active disturbance rejection control and the traditional high-order differential feedback control are adopted to test the tracking performance and disturbance rejection ability of the quadrotor unmanned aerial vehicle for different reference attitude. The results show that the proposed scheme is significantly superior to other ones in transient performance, steady-state tracking accuracy and robustness.

Key words: quadrotor unmanned aerial vehicle, attitude system, high-order differentiator, extended state observer, model-free control, unknown model function estimation, active disturbance rejection control